Game addiction

Studying data from Steam ¶

by Sebastian Wolf

Task¶

Your client is a mental health expert from an NGO who is interested in understanding more about gaming and the potentially addictive effect it can have on some individuals. You are meeting the client in a few days and they would like you to extract and present insights from the Steam dataset to help them in their research.

Please use whichever tools you feel the most comfortable with, but we do recommend Tableau which is a popular choice. Tableau is free for students and there is also a free trial available here.

My approach in these slides¶

Explore the context
- Geographic distribution of gamers
- Evolution of gamers on Steam
- Times when gamers are active
Analyze game addiction
- Define game addiction in the dataset
- Explore correlations of playing time with several engagement factors
- Derive conclusions and suggest focus for further analysis

Exploring the context¶

Geographic distribution of gamers
Evolution of gamers on steam
When gamers are active

The Steam dataset includes gamers from the entire world¶

Yet, most steam users report their location to be in the United States and Russia
Explore user counts per country by hovering over the map.

The number of new friendships is proportional to the number of new users
We see 2-3 times as many new friendships as new users per day

Gamers are more active on weekends and mornings¶

The figure shows the day of the week and the hour when gamers sign up for Steam
We see heightened activity from Saturday morning to Monday morning

Total number of new accounts created by day of week and hour_created
	0	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19	20	21	22	23
dayofweek_created
1	20,712	24,111	26,832	31,342	34,095	36,218	38,641	40,540	42,115	43,042	43,233	42,606	39,839	35,783	31,119	26,809	22,926	20,648	19,307	17,880	16,454	14,946	14,116	13,566
2	14,412	15,932	18,235	20,946	24,119	27,304	31,342	34,746	37,464	39,771	41,727	40,832	39,022	34,727	29,931	25,468	21,908	20,008	18,899	17,768	16,401	14,957	13,890	13,366
3	14,313	16,409	18,892	21,698	24,832	28,003	32,019	34,880	37,742	39,369	41,241	41,093	38,825	34,088	29,350	25,058	21,991	20,085	19,911	18,395	17,032	15,379	14,445	14,331
4	14,790	16,674	18,663	21,668	24,928	29,029	32,721	35,825	38,279	40,294	41,175	40,272	37,919	34,552	30,678	26,029	22,960	21,000	19,750	18,820	17,630	16,241	15,168	14,850
5	15,722	17,339	19,468	22,754	25,457	28,868	32,343	35,709	38,432	41,012	42,140	42,238	40,476	36,321	32,086	27,216	23,704	21,386	20,126	18,942	17,688	16,515	15,676	15,540
6	16,600	18,336	20,696	24,241	27,735	32,039	36,887	40,296	42,730	45,495	46,455	46,023	43,803	40,327	36,501	31,101	27,244	24,522	23,321	22,993	22,021	21,111	20,438	20,172
7	21,960	24,851	29,457	33,788	37,520	40,902	43,537	46,180	48,280	50,307	51,070	48,617	45,448	41,444	36,958	31,822	27,636	24,587	22,707	22,155	21,191	20,189	19,576	19,505

Analyzing game addiction¶

Define addiction for this analysis
Explore correlations with engagement factors in 5 hypothesis
Summarize findings
Derive conclusions and suggest focus for further analysis

Game addiction affects only a small fraction of gamers¶

The median playtime among active gamers (users that were active in the 2 weeks prior to data retrieval) was only 30min per day
Yet, 25% of active gamers play nearly 2h per day
We will focus on the 1% of all gamers that play more than 5h per day on average
To learn about game addicition, we will explore correlations of playtime with several engagement factors

Hypothesis 1: The more friends I have, the more I play¶

Heavy gamers have more friends than regular gamers: average = 40 vs 7 and median 20 vs 1
There is a positive, but weak correlation of 0.3 between playtime and the number of friends
However, among heavy gamers, the friend count does not correlate with playtime
Takeaway: Heavy gamers generally have many friends on the network. But the number of friends does not correlate with their behaviour at the margin.

Hypothesis 2: The more my friends play...¶

Heavy gamers' friends (if active) play more than regular gamers' friends (~2h vs 1h per day)
The same holds for the median friend (1h45min vs 10min)
For heavy gamers, playtime correlation with friends is only 0.03, vs 0.24 for the regular gamer
Takeaway: Friends seem to be an influence, but less so at the margin.

Hypothesis 3: The more my heaviest gamer friend plays...¶

Correlation of playtime with the playtime of the friend that plays the most (0.29) is stronger than with the playtime of the average friend (0.24)
But again, this correlation is weaker for the heavy gamers (0.05)
Takeaway: The distribution of behaviour among friends matters. The maximum appears more important than the average.

Hypothesis 4: The more groups I am part of...¶

Rather than friends, it could be the diversity of groups that keeps users engaged
Yet, group membership is uncommon: The median gamer is part of 0 groups vs 1 for heavy gamers
Correlation with playtime is very low
Takeaway: Groups seem to be less important than friends in understanding game addiction

Hypothesis 5: The more multiplayer games I play...¶

It would seem that multiplayer games are more addictive, especially given the correlation of playtime and friends
However, there is nearly no correlation between the fraction of multiplayer games played and playtime
Takeaway: Regular gamers alike heavy gamers play a mix of single and multiplayer games

What we learn from correlations¶

Several factors have been found to correlate with playtime:
- The number of friendships
- The playtime of friends
- The maximum playtime among friends
- The number of group memberships
Correlation is not causation: We don't know whether the above factors cause addiction
But we do know that these factors form important parts of heavy gamer's gaming experiences and need to be explored to understand addiction
We also found that these factors seem to play a lesser role in explaining why heavy gamers play an hour more or less (low correlations at the margin)

Correlations between playtime and engagement factors are lower for heavy gamers¶

Conclusions¶

Game addiction is rare among steam users (playtime distribution follows a power law)
The social factor (friends, groups) seems to be key to increased user engagement
To study addiction further, we need more information regarding gamers' interaction with games